Investigation of the Impact from IL-2, IL-7, and IL-15 on the Growth and Signaling of Activated CD4+ T Cells

Home , CD28, Cytokine, IL17RB, Interleukin 15, Interleukin 7, Memory T cell

International Journal of Molecular Sciences

Article Investigation of the Impact from IL-2, IL-7, and IL-15 on the Growth and Signaling of Activated CD4+ T Cells

1, 1, 2 2 1, Canaan Coppola † , Brooks Hopkins † , Steven Huhn , Zhimei Du , Zuyi Huang * and William J. Kelly 1,*

1 Department of Chemical Engineering, Villanova University, Villanova, PA 19085, USA; [email protected] (C.C.); [email protected] (B.H.) 2 Cell/Gene Therapy and Biologics Development, Merck & Co., Kenilworth, NJ 07033, USA; [email protected] (S.H.); [email protected] (Z.D.) * Correspondence: [email protected] (Z.H.); [email protected] (W.J.K.) These authors contributed equally to this work. † Received: 30 September 2020; Accepted: 19 October 2020; Published: 22 October 2020

Abstract: While CAR-T therapy is a growing and promising area of cancer research, it is limited by high cost and the difficulty of consistently culturing T-cells to therapeutically relevant concentrations ex-vivo. Cytokines IL-2, IL-7 and IL-15 have been found to stimulate the growth of T cells, however, the optimized combination of these three cytokines for T cell proliferation is unknown. In this study, we designed an integrated experimental and modeling approach to optimize cytokine supplementation for rapid expansion in clinical applications. We assessed the growth data for statistical improvements over no cytokine supplementation and used a systems biology approach to identify genes with the highest magnitude of expression change from control at several time points. Further, we developed a predictive mathematical model to project the growth rate for various cytokine combinations, and investigate genes and reactions regulated by cytokines in activated CD4+ T cells. The most favorable conditions from the T cell growth study and from the predictive model align to include the full range of IL-2 and IL-7 studied, and at lower levels of IL-15 (6 ng/mL or 36 ng/mL). The highest growth rates were observed where either IL-2 or IL-7 was at the highest concentration tested (15 ng/mL for IL-2 and 80 ng/mL for IL-7) while the other was at the lowest (1 ng/mL for IL-2 and 6 ng/mL for IL-7), or where both IL-2 and IL-7 concentrations are moderate-corresponding to condition keys 200, 020, and 110 respectively. This suggests a synergistic interaction of IL-2 and IL-7 with regards to promoting optimal proliferation and survival of the activated CD4+ T cells. Transcriptomic data analysis identified the genes and transcriptional regulators up/down-regulated by each of the cytokines IL-2, IL-7, and IL-15. It was found that the genes with persistent expressing changes were associated with major pathways involved in cell growth and proliferation. In addition to influencing T cell metabolism, the three cytokines were found to regulate specific genes involved in TCR, JAK/STAT, MAPK, AKT and PI3K-AKT signaling. The developed Fuzzy model that can predict the growth rate of activated CD4+ T cells for various combinations of cytokines, along with identified optimal cytokine cocktails and important genes found in transcriptomic data, can pave the way for optimizing activated CD4 T cells by regulating cytokines in the clinical setting.

Keywords: RNA sequencing; activated CD4+ T cell; interleukin-2; interleukin-7; interleukin-15

1. Introduction In immunotherapies, adoptive or engineered T cell therapies are a personalized approach to successful remediation of various diseases. Identifying a robust process that can eﬃciently induce proliferation of naïve T cells is essential in T cell manufacturing process development. In an article

Int. J. Mol. Sci. 2020, 21, 7814; doi:10.3390/ijms21217814 www.mdpi.com/journal/ijms Int. J. Mol. Sci. 2020, 21, 7814 2 of 23 published in 2019 by Cancer Discovery, the authors touch on the importance of naïve CD4+ T cells in pediatric cancer patients receiving immunotherapies. The effects of repetitive chemotherapy cycles deplete naïve T cells in many pediatric cancer patients, which minimizes the potential to induce proliferation successfully after adoptive cellular therapies [1]. Furthermore, the complete impacts of cytokines–and especially the combinations of cytokines–is poorly understood in the context of the rapid expansion of ex-vivo T-cell cultures and in how specific T-cell subsets differentially respond. As part of the CAR-T cancer therapy, transfected T-cells are grown ex-vivo until therapeutically relevant concentrations are reached. This process can be challenging to do efficiently, as the in-vivo environment normally provides the T-cells with growth-regulating signals that are absent in conventional medium formulations. The addition of the interleukin-2, 7, and 15 cytokines has been shown to induce human T-cell activation proliferation in-vitro, although the cytokine concentrations and growth rates are not linearly correlated [2,3]. There is extensive qualitative and less quantitative information on the merits of the cytokines’ impact on T cell proliferation individually per phenotype. There is also lacking evidence in an experimentally design study to reveal the optimal combination and overall impact the cytokines have on the proliferation of the CD4+ T cell phenotype [3]. It is useful, therefore, to study the impact of the concentration of the aforementioned cytokines on the proliferation and signal transduction of the naïve CD4+ T cells. Another aim of the study is to generate more data and information for naïve CD4+ T cells due to the inherent challenges this phenotype presents. Natural killer (NK) cells and cytolytic CD8+ T cells are often more easily influenced with cytokines into proliferation, but the CD4+ T cells are less responsive to one cytokine alone [3–7]. According to the literature, there is more to be done to unlock a useful cocktail of cytokines to balance the slow growth of the naïve CD4+ T cells and the tendency for this phenotype to undergo differentiation to a different phenotype [8]. CD4+ naïve T cells play critical roles in aiding the expansion of adoptive immunotherapies in vivo, and therefore this study can positively impact manufacturing productivity and efficacy while reducing uncertainty of cytokine concentration [9,10]. To develop the conditions for the experiment, the team studied historical reference points that may lead to the appropriate boundary conditions. As naïve CD4+ T cells are less responsive than naïve CD8+ T cells and NK cells [5], we would need to carefully examine the ranges of application for each of the cytokines. Alves, et al. describes IL-15 being used to stimulate growth of naïve CD4+ and naive CD8+ T cells, as well as NK cells [5]. In that experiment, naïve CD4+ T cells were required at least 40 ng/mL to divide at approximately 20% rate. This led roughly 40 ng/mL to become the median in the condition data set for IL-15. Geginat et al. note that in particular, naïve CD4+ T cells respond quickly to antigen-presenting molecules (APM), having only a 40-h lag prior to first division, when compared with cytokine stimulation alone having a lag of 72 h [3]. The division times are characteristically different in the naive CD4+ T cells when exposed to APMs versus cytokines, where the division times averaged 10 h when responding to APMs versus 30 h when responding to cytokines [3]. IL-7 and IL-15 cytokines were used in the study to induce proliferation, whereby the concentration was approximately 25 ng/mL for both cytokines in the study [3]. Another study was able to provide some background for concentrations of IL-2 needed for the study; Litvinova, et al. describes the concentration of IL-2 used to stimulate significant growth in CD4+ T cells as 1 ng/mL [11]. When compared to IL-7 and IL-15, IL-2 alone was the only cytokine to drive significant growth. Silva et al., describes the proliferation of naïve CD4+ T cell maintenance in patients with thymectomy, using IL-2 and IL-7 to stimulate the growth [12]. The concentration of IL-2 was 2 ng/mL and the IL-7 concentration was 10 ng/mL which successfully drove improved proliferation of the cells. Jaleco, et al, ran an experiment of naïve and memory CD4+ T cell subsets, attempting to walk the tight rope of proliferation and apoptosis [4]. The findings show that IL-2 and IL-7 act together for proliferative cell divisions, which is significantly more effective than using IL-2 or IL-7 alone. The concentration of IL-7 was again 10 ng/mL and the concentration of IL-2 was 6 ng/mL. Finally, Dooms and Abbas, as well as Kinter, et al., tested up to 100 ng/mL for IL-2, IL-7 Int. J. Mol. Sci. 2020, 21, 7814 3 of 23 and IL-15, but there were cell death implications of the highest concentration [13,14]. We thus deduce that the optimal concentration lies likely between 1–100 ng/mL for each of the cytokines. On the basis of the review on the cytokine concentrations used to promote T cell growth, sixteen conditions were designed in this study to investigate the expansion of the CD4+ T-cell activated with anti-CD3/CD28 dynabeads under the influence of various combinations of IL-2, IL-7 and IL-15 in a 96-well plate format. Three levels of concentrations, which represent low, moderate and high concentrations of each cytokine, were applied in four experiments. The growth data were thoroughly analyzed with ANOVA tests [15] to identify the cytokine cocktails that were able to significantly enhance the growth of activated CD4+ T cells when compared to the control condition. A Fuzzy model, which is known for its good prediction capability [16], was then developed to predict the growth of activated CD4+ T cells and study the interaction of cytokines in promoting activated CD4+ T cell growth. Finally, transcriptomic data was obtained for each cytokine stimulation and analyzed with principal component analysis [17], GeneMania [18], and MetaCore [19] to identify the genes/reactions regulated by each cytokines in activated CD4+ T cell. The genes whose profiles are correlated with the growth rates were also identified. The relevance of this work stems from its innate foundational-level exploration for the advancement of understanding the quantifiable impact of each of the three cytokines on the growth of activated CD4+ T cells. Additionally, we sought to study the synergistic impact of different combinations of cytokines on the growth rate of the T cells. To our knowledge, this is the first trial to systematically investigate the combination of the aforementioned cytokines to accelerate the growth of activated CD4+ T cells.

2. Results

2.1. Identify Optimal Cytokine Combinations from the Statistical Analysis of Cell Growth Rates Cells were thawed from cryopreservation in DMSO and activated with anti-CD3/CD28 dynabeads on day 0 of the experiment, and cytokine supplementation at the experimental conditions was performed on day three after media acclimation and post-activation anergy had completed. Pairwise t-tests were then conducted on the growth rates obtained from four datasets for 16 conditions each (Table1).

Table 1. Pairwise t-tests for the growth rates between the 16 conditions, using the growth rate data shown in Figure1. p-values are shown at the intersection of the row and column for each condition pair. The table was generated using R studio’s “pairwise.t.test” command, and the assumptions of normality and homogeneity of variances were verified using the Shapiro-Wilk and Bartlett tests, respectively. Statistical differences between means were identified for α < 0.05.

000 001 002 010 011 020 022 100 101 110 111 200 202 220 222 001 0.282 ------002 0.293 0.890 ------010 0.214 0.874 0.755 ------011 0.029 0.340 0.225 0.433 ------020 0.060 0.435 0.323 0.533 0.917 ------022 0.126 0.662 0.538 0.781 0.629 0.728 ------100 0.164 0.763 0.640 0.887 0.528 0.629 0.892 ------101 0.043 0.428 0.300 0.534 0.853 0.951 0.750 0.640 ------110 0.020 0.270 0.169 0.350 0.864 0.797 0.525 0.434 0.722 ------111 0.124 0.657 0.533 0.775 0.635 0.734 0.994 0.886 0.756 0.531 ----- 200 0.035 0.376 0.255 0.474 0.937 0.974 0.679 0.575 0.916 0.802 0.685 ---- 202 0.075 0.493 0.376 0.597 0.835 0.923 0.801 0.699 0.965 0.717 0.807 0.890 --- 220 0.584 0.551 0.608 0.443 0.090 0.154 0.287 0.358 0.128 0.064 0.283 0.105 0.184 -- 222 0.240 0.925 0.809 0.948 0.393 0.492 0.732 0.836 0.489 0.316 0.726 0.432 0.553 0.485 - no cyt 0.878 0.227 0.231 0.170 0.021 0.046 0.098 0.129 0.031 0.014 0.096 0.025 0.057 0.485 0.192 Int. J. Mol. Sci. 2020, 21, 7814 4 of 23

Int. J. Mol. Sci. 2020, 21, x FOR PEER REVIEW 4 of 23 The growth rates were normalized to the average growth rate for each data set for easy comparison ofconditions improved 200, and 020, inhibited 110, and growth, 011 are and statistically the data show high that when the growth compared rates to for the conditions negative 200, control 020, 110,(no andcytokine 011 aresupplementation) statistically high and when baseline compared 000 condition to the negative (Figure control 1). (no cytokine supplementation) and baseline 000 condition (Figure1).

(A) (B)

Figure 1. Various combinations of the IL-2, IL-7, and IL-15 cytokines differentially differentially regulate activated CD4+ lymphocyte proliferation:proliferation: ( A(A) Mean) Mean growth growth rate rate data data were were obtained obtained in four in data four sets, data with sets, growth with ratesgrowth calculated rates calculated for a four-day for a four-day period period after post-thawafter post-thaw acclimation, acclimation, and and each each set normalizedset normalized to itsto respectiveits respective average average growth growth rate. rate. Error Error bars bars represent represent ±1 1 SE, SE, and and groupinggrouping lettersletters showshow statistically ± significantsignificant differences differences ( (αα << 0.05)0.05) as asmeasured measured by bypairwise pairwise t-tests.t-tests. p-valuesp-values are shown are shown in Table in 1. Table Note:1. Note:“a” in “a”the infigure the figureindicates indicates conditions conditions with growth with growth rates significantly rates significantly higher than higher the than negative the negative control controland the andbaseline the baseline 000 conditions; 000 conditions; and “b” and represents “b” represents the negative the negative control control and 000 and conditions 000 conditions (see Table (see Table5 for 5experimental for experimental concentrations). concentrations). (B) (ExampleB) Example growth growth profile profile of of selected selected conditions, conditions, where cytokines werewere added added on on day day 3 and 3 and growths growths rates rates calculated calculated between between days 3 anddays 7. 3 Error and bars7. Error represent bars represent1 SD. Full ± 1 data SD. setsFull aredata available sets are uponavailable request. upon request. ±

Since the initially naïve cells will produce endogenousendogenous cytokines, the no cytokine control was used as aa baselinebaseline forfor thethe eeffectsﬀects of of interleukin interleukin supplementation. supplementation. The The similarity similarity of of the the 000 000 condition condition to theto the no no cytokine cytokine control control suggests suggests a critical a critical level level of supplementationof supplementation for for initial initial proliferation proliferation that that was was not metnot met by the by lowthe levelslow levels of the of 000 the condition.000 condition. The conditionThe condition keys keys represent represent the concentrations the concentrations of each of cytokineeach cytokine (IL-2, (IL-2, IL-7, IL-15)IL-7, IL-15) at the threeat the dithreeﬀerent different experimental experimental levels, withlevels, 0, 1,with and 0, 2 1, in and each 2 positionin each indicatingposition indicating the low, medium,the low, medium, and high and concentrations, high concentrations, respectively respectively (Table2). (Table 2).

Table 2. The keys and concentrations of IL-2, IL-7 and IL-15 used in the experiments. Experimental Table 2. The keys and concentrations of IL-2, IL-7 and IL-15 used in the experiments. Experimental concentrations were chosen from literature precedent and reﬁned from initial growth studies and concentrations were chosen from literature precedent and refined from initial growth studies and modeling results. modeling results. Key IL-2 (ng/mL) IL-7 (ng/mL) IL-15 (ng/mL) Key IL-2 (ng/mL) IL-7 (ng/mL) IL-15 (ng/mL) PCPC 0 0 0 0 0 0,0,0 1 6 6 0,0,0 1 6 6 1,0,0 10 6 6 2,0,01,0,0 10 15 6 6 6 6 0,1,02,0,0 15 1 6 36 6 6 0,2,00,1,0 1 36 80 6 6 0,0,10,2,0 1 80 6 6 36 0,0,2 1 6 80 1,1,00,0,1 101 6 36 36 6 1,0,10,0,2 101 6 6 80 36 0,1,11,1,0 10 1 36 36 6 36 2,2,01,0,1 10 15 6 80 36 6 2,0,20,1,1 151 36 6 36 80 0,2,2 1 80 80 1,1,12,2,0 15 10 80 36 6 36 2,2,22,0,2 15 6 80 80 80 0,2,2 1 80 80 1,1,1 10 36 36 2,2,2 15 80 80

Int. J. Mol. Sci. 2020, 21, 7814 5 of 23

2.2. Investigate the Interaction of the Three Cytokines in Promoting the Growth of CD4 + Naïve T Cells Factorial analysis of variance on the combined data showed a significant interaction between all three cytokines and stronger interaction of IL-2 and IL-7 specifically (Table3). All three of these cytokines signal through the common γ chain and have been shown independently to differentially regulate the proliferation of T cell subsets. The interaction of IL-2 and IL-7 is also present and significant in three of the four data sets alone, supporting previous findings of the importance and synergy of these cytokines for the proliferation and survival of naïve CD4+ lymphocytes. Co-stimulation with IL-2 and IL-7 has been implicated in modulating the proliferative, apoptotic, and necrotic susceptibilities of naïve T cells [4]. IL-7 alone can promote the survival of naïve CD4+ T cells ex vivo, while the combination also increases their sensitivity to Fas-mediated cell death. We hypothesize that higher concentrations of these cytokines may allow the cell death mechanism to begin to dominate, while moderate concentrations allow for faster proliferation after Ag activation without triggering extensive cell death.

Table 3. Statistical analysis of interaction of IL-2, IL-7, and IL-15 in influencing activated CD4+ T-cell growth rates by factorial analysis of variance (ANOVA). The analysis was conducted with R Studio’s “aov” command, and normality of residuals and homogeneity of variances were verified using the Shapiro and Bartlett tests, respectively. Pr (>F) of less than 0.05 was used to test for significance of interaction between interleukins. Significant interaction between IL-2 and IL-7 indicates that the effect of IL-2 is not consistent across all levels of IL-7, and the converse, suggesting an interactive effect of both present.

Df Sum Sq Mean Sq F Value Pr (>F) IL-2 1 0.049 0.0486 0.533 0.4688 IL-7 1 0.030 0.0303 0.333 0.5667 IL-15 1 0.012 0.0124 0.136 0.7137 IL-2:IL-7 1 0.618 0.6180 6.785 0.0122 IL-2:IL-15 1 0.231 0.2305 2.531 0.1182 IL-7:IL-15 1 0.003 0.0030 0.033 0.8560 IL-2:IL-7:IL-15 1 0.412 0.4122 4.525 0.0386 Residuals 48 4.372 0.0911

IL-15 has also been implicated in the proliferation and survival of T-cell populations, notably through its interaction with the IL-2 receptor [18]. Our growth study suggests that moderate IL-15 concentrations can improve the seven-day growth rate from thaw over similar conditions with high IL-15 concentrations; this may be due to competitive binding of IL-15 with IL-2 through their common IL-2 receptor, with high IL-15 concentration reducing the proliferative impact of IL-2. These three cytokines are regulators of the relative amounts of T-cell subsets in a mixed T-cell population [4], and therefore are capable of sensitizing cells to death-inducing pathways, or to proliferative pathways. In our study of Ag-activated CD4+ cells, over-stimulation by IL-15 tended to reduce growth rates, while the cells were more tolerant to any level of IL-2 and IL-7 investigated in this study. High IL-2 and low IL-7 and IL-15 (condition 200) gave the most rapid expansion at the cost of overall viability, while the 011 and 110 conditions nearly as rapidly and had higher viabilities (viability data not shown).

2.3. Develop a Fuzzy Model to Predict the Growth Rates of Activated CD4+ T Cells for Various Cytokine Combinations One experimental dataset (obtained in September 2019) was used to estimate the parameters in the Fuzzy model (Figure2A). The cytokine concentrations from another experimental dataset (obtained in October 2019) for eight new conditions were fed as the inputs of the developed Fuzzy model to predict the growth rate (Figure2B). Int.Int. J. J. Mol. Mol. Sci. Sci.2020 2020, 21, 21, 7814, x FOR PEER REVIEW 66 of of 23 23

(A) (B)

FigureFigure 2. 2.Fuzzy Fuzzy model model prediction prediction with experimentalwith experimental data from data September from September (A); Fuzzy ( ModelA); Fuzzy prediction Model withprediction experimental with experimental data from Octoberdata from (B October). Experimental (B). Experimental data in (A data) is in from (A) ais singlefrom a dataset single dataset used toused estimate to estimate parameters parameters for the for fuzzy the fu model,zzy model, and and model model predictions predictions in (inB) (B were) were tested tested in in a a second second experimentalexperimental run run to to validate validate the the model model predictions predictions and and analyze analyze the the choice choice of of cytokine cytokine concentrations concentrations forfor future future experiments. experiments. MATLAB MATLAB was was used used for for these these modeling modeling studies. studies.

ThoseThose eight eight new new conditions, conditions, which which were were not included not included in the 16in designedthe 16 designed conditions, conditions, were selected were fromselected the fuzzyfrom the model fuzzy to model evaluate to evaluate the model the for model the conditionsfor the conditions returning returning high, moderatehigh, moderate and low and growthlow growth rates, rates, respectively. respectively. Amid Amid the inconsistencies the inconsistencies in the in above the above figure figure can be can seen, be theseen, model the model was testedwas tested against against a different a different set of conditions set of conditions that had that diff erenthad different levels of levels cytokines of cytokines compared compared to the original to the dataoriginal sets. data The sets. objective The objective was to validate was to thevalidate spectrum the sp ofectrum levels of that levels could that be could seen inbe aseen singular in a singular test of thetest integrity of the integrity of the 3-D of scatterthe 3-D heat scatter map. heat The map. model The does model show does some show consistency some consistency with the with data the in thisdata case.in this There case. is There truly onlyis truly one only data one point data that point is inconsistent that is inconsistent from the experimentalfrom the experimental data. The data. point The of cytokinepoint of levels cytokine 1, 6, levels 60 ng /1,mL 6, for60 IL-2,ng/mL IL-7 for and IL-2, IL-15, IL-7 respectively, and IL-15, respectively, most closely most resembles closely the resembles original conditionthe original 0,0,2, condition which in 0,0,2, data which set I was in data one ofset the I was lowest one performing of the lowest conditions performing (as seenconditions in Figure (as2 seenA). Thein Figure data shown 2A). The in Figuredata shown2B (i.e., inthe Figure October 2B (i.e., 2019 the data) October was 2019 not included data) was in no Figuret included1. Only in Figure one of 1. theOnly other one three of the datasets other three (i.e., datase Septemberts (i.e., 2019 September data) was 2019 used data) in Fuzzywas used modeling in Fuzzy (Figure modeling2A), as(Figure we aimed2A), as to we develop aimed the to model develop to validatethe model our to choice validate of the our cytokine choice of concentrations the cytokine andconcentrations investigate theand interactioninvestigate of the the interaction three cytokines of the before three cytokines more experiments before more were experiments conducted. were conducted. SinceSince the the developed developed Fuzzy Fuzzy model model has has been been validated validated by by experimental experimental data, data, it it is is thus thus of of value value to to + predictpredict the the growth growth rates rates of of activated activated CD4 CD4+T T cell cell growth growth rates rates for for various various combination combination of of IL-2, IL-2, IL-7 IL-7 andand IL-15. IL-15. A A 3-D 3-D surface surface heat heat map map was was generated generated in in Figure Figure3 from3 from the the Fuzzy Fuzzy model model by by evaluating evaluating all all possiblepossible combinations combinations of of IL-2 IL-2 (0–40 (0–40 ng ng/mL),/mL), IL-7 IL-7 (0–100 (0–100 ng ng/mL),/mL), and and IL-15 IL-15 (0–100 (0–100 ng ng/mL)./mL). Consistent Consistent withwith what what was was observed observed in Sections in Sections 2.1 and 2.22.1, theand Fuzzy 2.2, modelthe Fuzzy predicts model that moderatepredicts concentrationsthat moderate + ofconcentrations the three cytokines of the canthree return cytokines acceptable can retu growthrn acceptable rates of activatedgrowth rates CD4 of activatedT cells (as CD4 shown+ T cells in the (as middleshownleft in the and middle right of left the and heat right map). of the As impliedheat map). by As the implied data shown by the in data Figure shown2A, the in FuzzyFigure model 2A, the returnsFuzzy highmodel growth returns rate high for growth high concentrations rate for high concen of thetrations cocktails of IL-7~IL-15 the cocktails (the IL-7~IL-15 left top of (the Figure left 3top) andof Figure IL-2~IL-15 3) and (the IL-2~IL-15 right top (the of Figure right3 top). Figure of Figure3 also 3). indicates Figure 3 strong also indicates synergistic strong e ff ectsynergistic between effect IL-7 andbetween IL-15 IL-7 and betweenand IL-15 IL-2 and and between IL-15. IL-2 Additional and IL-15. synergistic Additional eff ectsynergistic was also effect observed was also for IL-2observed and IL-7for inIL-2 Figure and 3IL-7. This in Figure is also consistent3. This is also with consistent the result with shown the in result the previous shown in section. the previous section.

Int. J. Mol. Sci. 2020, 21, 7814 7 of 23 Int. J. Mol. Sci. 2020, 21, x FOR PEER REVIEW 7 of 23

+ 1 FigureFigure 3. 3.The The growth growth rates rates of activatedof activated CD4 CD4T-cells+ T-cells (hr −(hr)−1 predicted) predicted by by the the model model for for di ff differenterent combinationscombinations of cytokines.of cytokines. The The model model was was assessed assessed over over the the entire entire concentration concentration space space investigated investigated by the growth studies to generate this map, with all possible combinations of IL-2 [0–40 ng/mL], by the growth studies to generate this map, with all possible combinations of IL∈-2 ∈ [0–40 ng/mL], IL-7 [0–100 ng/mL], and IL-15 [0–100 ng/mL]. The growth rate scale in (hr 1) are shown on the IL∈-7 ∈ [0–100 ng/mL], and IL∈-15 ∈ [0–100 ng/mL]. The growth rate scale in− (hr−1) are shown on the right, with yellow representing fastest proliferation and dark blue the lowest. right, with yellow representing fastest proliferation and dark blue the lowest. 2.4. Investigate Genes Indicating the Purtabation in the Metabolism of Activated CD4+ T Cells Stimulated by Individual2.4. Investigate Cytokines Genes Indicating the Purtabation in the Metabolism of Activated CD4+ T Cells Stimulated by Individual Cytokines The expression variation of genes that are involved in positive proliferation and negative proliferationThe expression were identified variation in Table of genes4 using that the are program involved Metacore, in positive along proliferation with overlapping and negative the upproliferation/down-regulated were genes identified with severalin Table databases 4 using (such the program as Biological Metacore Magnetic, along Resonance with overlapping Data Bank the andup/down Mammalian-regulated Metabolic genesEnzyme with several Database). databases One (such of the as moreBiological important Magnetic aspects Resonance we monitored Data Bank wasand determining Mammalian what Metabolic metabolic Enzyme pathways Database were). One persistently of the more upregulated important aspects with high-magnitude we monitored was fold-increasedetermining from what baseline metabolic genes pathways represented. were Using persistently genes from upregulated the glycolysis, with high inositol-magnitude phosphate, fold- glutamineincrease metabolism, from baseline PI3K-AKT, genes represented.mTOR, Myc, TCA Using cycle, genes protein from synthesis the glycolysis, and glycerophospholipid inositol phosphate, pathwaysglutamine that are metabolism, well-known forPI3K their-AKT, important mTOR, roles Myc, in regulating TCA cell cycle, metabolism, protein we synthesis analyzed the and genesglycerophospholipid which have ties to pathways these pathways, that are specifically well-known which for are their up important and down-regulated roles in regulating by each of cell IL-2,metabolism IL-7 and IL-15., we analyzed the genes which have ties to these pathways, specifically which are up and downGenes-regulat associateded by witheach IL-2of IL stimulation-2, IL-7 and resultingIL-15. in gene upregulation from the aforementioned metabolicGenes gene associated pathways with include IL-2 genes stimulation associated resulting with thein gene metabolism upregulation of lipids, from carbohydrates the aforementioned and proteins,metabolic including gene pathwaysAPOE, PFKFB1 include, IGFBP2 genes, associatedRPS9 and PGAM4 with the. Consequently, metabolism of one lipids, gene ca downregulatedrbohydrates and viaproteins IL-2 stimulation, including was APOECHST1, ,PFKFB1 involved, IGFBP2 in amino, RPS9 acid metabolism. and PGAM4 Genes. Consequently, upregulated via one IL-7 gene stimulationdownregulated include via genes IL-2 involved stimulation in carbohydrate, was CHST1, lipid, involved protein in andamino amino acid acid metabolism. metabolism Genes as well:upregulatedAPOE, PFKFB1 via IL,-7CES3 stimulation, IGFBP2 include, DMGDH genesand involvedKLK2. Genesin carbohydrate, associated lipid, with IL-7protein stimulation and amino resultingacid metabolism in downregulation as well: includeAPOE, CHST13PFKFB1, CHST1CES3, IGFBP2, MGIL,, and DMGDHACSM1 and. Finally KLK2 genes. Genes associated associated with with IL-15IL-7 upregulation stimulation areresulting focused in indownregulation mostly in lipid include metabolism: CHST13ACHE, CHST1, APOE, MGIL, RDH5, and, PRKAA2 ACSM1, CDS1. Finally, andgenesDMGDH associated. Those with genes IL- downregulated15 upregulation by are IL-15 focused include in mostlyMRM1 in, ENPP lipid andmetabolism:MT5C1A ACHE. Key genes, APOE, fromRDH5 these, PRKAA2 lists are, compiled CDS1, and in TableDMGDH4. . Those genes downregulated by IL-15 include MRM1, ENPP and MT5C1A. Key genes from these lists are compiled in Table 4.

Int. J. Mol. Sci. 2020, 21, 7814 8 of 23

Table 4. The functions of metabolic genes involved in positive and negative proliferation that have large change in their expression levels. Data for up- and down-regulated genes were obtained from RNA sequencing of samples supplemented at the no cytokine, 000, 200, 020, and 002 conditions. Genes were identiﬁed using the program Metacore and referenced with several databases, including the Biological Magnetic Resonance Data Bank and the Mammalian Metabolic Enzyme Database.

Genes for Positive Genes for Negative Functions of Genes Functions of Genes Proliferation Proliferation PFKFB1 Carbohydrate metabolism KIFC3 Metabolism of Proteins APOE Lipid metabolism MGLL Lipid metabolism PLA2G6 Lipid metabolism CHST1 Amino Acid Metabolism TPTE2 Lipid metabolism CDO1 Amino Acid Metabolism OLAH Citrate & Glyoxylate cycle MRM1 RNA Metabolism ACHE Lipid Metabolism AICDA Nucleotide Metabolism BAAT Steroid Metabolism SMPD3 Lipid Metabolism LDHC Amino Acid Metabolism VEGFC Growth Factor RDH5 Lipid metabolism PTHLH Hormone ACAD11 Enzyme SFRP2 Wnt Signaling PRKAA2 Lipid Metabolism CES3 Lipid Metabolism DMGDH Lipid Metabolism EGR4 Growth Factor KLK2 Protein metabolism TSHR Thyroid cell metabolism TGFB2 Growth Factor

In order to elucidate the functional differences between cytokines and their response to cell growth, we next analyzed significantly changed transcripts associated with proliferation related ontology vs. the control condition (Supplementary Figure S1). All cytokines shared commonalities via induction of HILPDA, a gene which enhances cell growth via lipid metabolism and potent downregulation of SFRP1, a gene involved in delay of S phase. Detailed analysis identified a common set of factors closely shared between IL-2 and IL-7. These shared genes revealed upregulation to PGDFB, JAK/STAT (EPOR), and Beta-catenin (PTPRU) related signaling as well as dramatic negative changes to angiogenic factors (VEGFC, FGF18, FGF2). Conversely, IL-17 resulted in a dramatic increase of TGFB2, downregulation of ERK signaling (GAREM2), and resulted in a decrease in expression of transcriptional programmer, CHD5.

2.5. Study Genes Indicating the Purtabation in the Signal Transduction of Activated CD4+ T Cells Stimulated by Individual Cytokines The genes with 50% change in their expression levels when compared to the control condition ± were identiﬁed via the approaches shown in the Methods section. Since hundreds of genes were up/down-regulated by each of IL-2, IL-7 and IL-15, principal component analysis was implemented to project those genes according to their expression time-proﬁles onto a two-dimensional space. The outlier genes were analyzed to identify the reactions unique to each cytokine, while the bulk genes were further analyzed by the program Metacore to identify the gene-regulatory networks induced by each of the three cytokines. Int. J. Mol. Sci. 2020, 21, 7814 9 of 23

Int. J. Mol. Sci. 2020, 21, x FOR PEER REVIEW 9 of 23 2.5.1. Outlier Gene Analysis 2.5.1. Outlier Gene Analysis Outlier genes were identified by principal component analysis and clustering from the Outlier genes were identified by principal component analysis and clustering from the transcriptomic data for activated CD4+ T cells stimulated by IL-2, IL-7, and IL-15, respectively transcriptomic data for activated CD4+ T cells stimulated by IL-2, IL-7, and IL-15, respectively (Figure4A,C,E). The functions of those outlier genes were identified from the program GeneMANIA (Figures 4A,C,E). The functions of those outlier genes were identified from the program GeneMANIA and listed in Figure4B,D,F. In Figure4B, the gene functions listed represent unique outlier genes from and listed in Figures 4B,D,F. In Figure 4B, the gene functions listed represent unique outlier genes the PCA study of genes with significant expression change via stimulation by IL-2. The gene functions from the PCA study of genes with significant expression change via stimulation by IL-2. The gene of the genes stimulated by IL-2 seem to be related to cell and leukocyte chemotaxis, chemokine receptor functions of the genes stimulated by IL-2 seem to be related to cell and leukocyte chemotaxis, activity, cellular defense response and cell motility. In Figure4D, the gene functions listed based on the chemokine receptor activity, cellular defense response and cell motility. In Figure 4D, the gene outlier genes from IL-7 stimulation, which include genes involved in chloride transport, cell chemotaxis, functions listed based on the outlier genes from IL-7 stimulation, which include genes involved in chloride transmembrane activity, and cytokine receptor activity. Finally, in Figure4F, the gene functions chloride transport, cell chemotaxis, chloride transmembrane activity, and cytokine receptor activity. listed based on the outlier genes from IL-15 stimulation include genes with functions, including cell Finally, in Figure 4F, the gene functions listed based on the outlier genes from IL-15 stimulation growth, zinc ion responses, cellular response to metal ions, and regulation of calcium ion transport include genes with functions, including cell growth, zinc ion responses, cellular response to metal into the cell. These gene functions are consistent with the types of genes which are outliers in the ions, and regulation of calcium ion transport into the cell. These gene functions are consistent with PCA study. the types of genes which are outliers in the PCA study.

Genes in Function Ne twork receptor-mediated endocytosis 10 chemokine receptor activity 9 G-protein coupled chemoattractant receptor activity 8 cell chemotaxis 8 regulation of chemotaxis 7 cholesterol homeostasis 7 sterol homeostasis 7 regulation of behavior 7 leukocyte chemotaxis 7 cytokine receptor activity 7 plasma lipoprotein particle clearance 7 positive regulation of chemotaxis 7 chemokine binding 6 G-protein coupled peptide receptor activity 6 lipid home ostasis 6 chemokine-mediated signaling pathway 6 positive regulation of locomotion 6 positive regulation of behavior 6 peptide receptor activity 6 behavior 6 regulation of plasma lipoprotein particle levels 6 myeloid leukocyte migration 6 positive regulation of positive chemotaxis 6 regulation of positive chemotaxis 6 receptor metabolic process 6 re ce ptor inte rnalization 6 regulation of leukocyte chemotaxis 6 cellular defense response 6 positive regulation of cell migration 6 positive re gulation of ce ll motility 6 (A) (B)

Figure 4. Cont Genes with significant expression chanage in CD4 naive stimulated by IL7 . 5 Genes in

CXCL8 4 SIX4 Function Ne twork TMEM213 PAX2 chloride channel activity 3 3 SNX31 DACT2 CACNA1E LOC729652 anion channel activity 3 2 CLCN4 HSF4 receptor metabolic proce ss 4 ADM 1 LINC00346 chloride tramsmembrane transporter activity 3 IL22 ARRDC3 PPDPF 0 SNORD116-20 chloride transmembrane transport 3 DPYSL5 SNORD95 SNORD70B SNORD96A SNORD75 -1 receptor internalization 3 SNORD47 Second Principal Component Principal Second SFRP2 chloride transport 3 -2 FRRS1L TMEM221 ETNPPL CEND1 G-protein coupled receptor binding 4 -3 MIR4637 ITGB8 cytokine receptor activity 3

-4 ACKR1 cell chemotaxis 4

CXCL9 -5 -15 -10 -5 0 5 10 15 First Principal Component

Int. J. Mol. Sci. 2020, 21, x FOR PEER REVIEW 9 of 23

2.5.1. Outlier Gene Analysis Outlier genes were identified by principal component analysis and clustering from the transcriptomic data for activated CD4+ T cells stimulated by IL-2, IL-7, and IL-15, respectively (Figures 4A,C,E). The functions of those outlier genes were identified from the program GeneMANIA and listed in Figures 4B,D,F. In Figure 4B, the gene functions listed represent unique outlier genes from the PCA study of genes with significant expression change via stimulation by IL-2. The gene functions of the genes stimulated by IL-2 seem to be related to cell and leukocyte chemotaxis, chemokine receptor activity, cellular defense response and cell motility. In Figure 4D, the gene functions listed based on the outlier genes from IL-7 stimulation, which include genes involved in chloride transport, cell chemotaxis, chloride transmembrane activity, and cytokine receptor activity. Finally, in Figure 4F, the gene functions listed based on the outlier genes from IL-15 stimulation include genes with functions, including cell growth, zinc ion responses, cellular response to metal ions, and regulation of calcium ion transport into the cell. These gene functions are consistent with the types of genes which are outliers in the PCA study.

Genes with significant expression chanage in CD4 naive stimulated by IL7 5 Genes in

-4 ACKR1 cell chemotaxis 4

CXCL9 -5 -15 -10 -5 0 5 10 15 First Principal Component Int. J. Mol. Sci. 2020, 21, x FOR PEER REVIEW 10 of 23 (C) (D) Genes in Function Ne twork cellular response to zinc ion 8 response to zinc ion 8 response to transition metal metal nanoparticle 8 re sponse to cadmium ion 6 ce llula r re sponse to me tal ion 8 ce llular re sponse to inorganic substance 8 negative regulation of growth 9 response to metal ion 8 perinuclear region of cytoplasm 10 regulation of growth 10 response to inorganic substance 8 zinc ion binding 8 positive re gula tion of ca lcium ion transport into cytosol 3

(E) (F)

FigureFigure 4. 4.PrincipalPrincipal component component analysis analysis (PCA) (PCA) projections projections of ofgenes genes with with significant signiﬁcant expression expression change change byby IL-2 IL-2 (A (),A IL-7(C)), IL-7(C) and and IL-15 IL-15 (E) (stimulation,E) stimulation, with with corresponding corresponding and respective and respective gene genefunction function of IL- of 2 (IL-2B), IL-7 (B), ( IL-7D), and (D), IL-15 and IL-15 (F) in ( Factivated) in activated CD4+ CD4T cells.+ T IL2 cells. mainly IL2 mainly regulated regulated genes genesrelated related to cell toand cell leukocyteand leukocyte chemotaxis, chemotaxis, chemokine chemokine receptor receptor activity, activity, cellular cellulardefense defenseresponse response and cell andmotility, cell motility,while IL7while regulated IL7 regulated genes involved genes involvedin chloride in transport, chloride transport, cell chemotaxis, cell chemotaxis, chloride transmembrane chloride transmembrane activity, andactivity, cytokine and receptor cytokine activity. receptor The activity. genes Theregulated genes by regulated IL15 were by involved IL15 were in involved cell growth, in cell zinc growth, ion responses,zinc ionresponses, cellular response cellular to response metal ions, to metal and ions,regulation and regulation of calcium of ion calcium transport ion transport into the intocell. the cell.

AsAs shown shown in inFigures Figure 4A,B,4A,B, the the genes genes with with respect respect to to IL-2 IL-2 stimulation stimulation include: include: CCL2CCL2, APOE, APOE, MB, MB, , CXCL9CXCL9, IL8, IL8, and, and ACKR1ACKR1. In. Inparticular, particular, CCL2CCL2, CXCL9, CXCL9 andand CXCL8/IL8CXCL8/IL8 areare all all markers markers of of proliferating proliferating cells.cells. Some Some of ofthe the standout standout genes genes include include MBMB (myoglobin),(myoglobin), as as well well as as ACKR1ACKR1, which, which is isa chemokine a chemokine receptor.receptor. It Itseems seems that that most most genes genes stimulated stimulated by by IL-2 IL-2 appear appear to to be be associated associated with with inflammation inflammation processes,processes, but but also also growth growth pathways pathways and and triggers triggers fo forr proliferation. proliferation. Outlier Outlier genes genes associated associated with with IL- IL-7 7 include: CXCL8CXCL8, CXCL9, IL-22, CACNA1E,, ADM,, SFRP2SFRP2 andand ITGB8ITGB8.. AlthoughAlthough therethere areare somesome genesgenes in inthis this list list which which are are involved involved in inflammation in inflammation and areand directing are directing agents, agents, there are there also are genes also associated genes associatedheavily with heavily ion with transport ion transport from the from cell withthe ce somell with trans-membrane some trans-membrane proteins proteins and channel and channel proteins. proteins.Outlier Outlier genes associated genes associated with IL-15 with include IL-15 in manyclude metal-ion many metal-ion associated associated proteins proteins like MT1H like, MT1H and also, andsignaling also signaling for other for cytokines other cytokines like IL17RB like .IL17RB. WhileWhile Figure Figure 44 showsshows the the genes genes regulated regulated by by individual individual cytokines, cytokines, these these outlier outlier genes genes are are commonlycommonly involved involved in in T-cell T-cell receptor receptor (TCR), (TCR), a a complex complex of of integral integral membrane proteins onon thethe surfacesurface of ofT T cells cells that that takes takes part part in thein the activation activation of T-cells of T-cells in response in response to an to antigen. an antigen. Genes Genes from pathwayfrom pathway analysis analysisthrough through NCI illustrate NCI illustrate the complexity the complexity of TCR signaling, of TCR whichsignaling, leads which to proliferation leads to proliferation and differentiation and + differentiationof activated CD4of activatedT cells [CD420].+ InT particular,cells [20]. In the particular, outlier genes the shownoutlier ingenes Figure shown4, especially in Figure those 4, especiallyregulated those by IL-2 regulated and IL-7, by wereIL-2 and found IL-7, to bewere related found to to related be related to TCR to signaling,related to asTCR shown signaling, in Table as5 . shown in Table 5. It is known that TCR signaling is an important step in activation of activated T cells in order to induce the proliferation and differentiation pathways [20]. The genes being produced in Table 5 as identified in the PCA are relevant as they are connected to the genes associated with direct TCR signaling. For example, CXCL8 is an important chemokine as a directing agent, involved in chemotaxis [21]. VEGFC is a growth factor connected to the MAPK and PI3K-AKT signaling pathways, which can lead to proliferation of T cells [22]. Unfortunately, it is also associated with tumor immune escape in certain cancers [22]. In another article involving VEGFC signaling, blocking VEGFC inhibits infiltration of CD4+ T cells, but does not affect T cell cytokine responses in the epithelium of rats burdened with obliterative airway disease (OAD), which could suggest the ability of VEGFC to act as a marker or directing agent for CD4+ T cells [23].

Int. J. Mol. Sci. 2020, 21, 7814 11 of 23

It is known that TCR signaling is an important step in activation of activated T cells in order to induce the proliferation and differentiation pathways [20]. The genes being produced in Table5 as identified in the PCA are relevant as they are connected to the genes associated with direct TCR signaling. For example, CXCL8 is an important chemokine as a directing agent, involved in chemotaxis [21]. VEGFC is a growth factor connected to the MAPK and PI3K-AKT signaling pathways, which can lead to proliferation of T cells [22]. Unfortunately, it is also associated with tumor immune escape in certain cancers [22]. In another article involving VEGFC signaling, blocking VEGFC inhibits infiltration of CD4+ T cells, but does not affect T cell cytokine responses in the epithelium of rats burdened with obliterative airway disease (OAD), which could suggest the ability of VEGFC to act as a marker or directing agent for CD4+ T cells [23].

Table 5. The outlier genes shown in Figure4 that are involved in TCR signaling. The data were generated through RNA sequencing of samples for the no cytokine, 000, 200, 020, and 002 conditions, and the listed genes were identiﬁed from that set through principal component analysis.

Cytokine Stimulation Outlier Gene from PCA CD4+ Naïve TCR Gene IL-2 CXCL8 PIK3CA IL-2 CXCL8 SHC1 IL-2 CXCL8 SLA2 IL-2 CXCL8 SOS1 IL-2 VEGFC PIK3CA IL-2 VEGFC SHC1 IL-2 VEGFC SOS1 IL-7 CACNA1E RASGRP2 IL-7 CXCL8 PIK3CA IL-7 CXCL8 SHC1 IL-7 DPYSL5 CRMP1 IL-7 CXCL8 SLA2 IL-7 CXCL8 SOS1 IL-7 ITGB8 FLNA

In addition to the outlier genes, the bulk genes not speciﬁc in Figure4 are further investigated to derive the gene regulatory networks stimulated by each of the three cytokines.

2.5.2. Bulk Gene Analysis Next, in order to further illustrate the effects of these cytokines on alternative pathways, we utilized the Metacore program to the bulk genes shown in Figure4 to calculate the most statistically a ffected genetic networks either shared (Supplementary Figure S2A) or uniquely changed (Supplementary Figure S2B) between conditions. Common amongst all was a strong angiogenic response, in addition to changes to cell morphology, adhesive properties, as well as EMT remodeling. Alternatively, the differentially expressed networks revealed a clustering of common components to IL-2 and IL-7 stimulation but not IL-15. This included the anti-apoptotic response, JAK/STAT induction, and alteration to chromatin. Conversely, these changes were largely absent from IL-15 signaling. In order to further describe these networks, we then generated process networks from the differentially expressed genes and visualized only the nodes demonstrating interconnection (Figure5). The IL-2 network (Figure5A) demonstrated a network clustered through the oncogene Src, general GPCR signaling, and the transcription factor AP-1. Change in chromatin modelling was noted via a variety of histone deacetylases. Several other transcription factors and molecules which are involved in IL-2 metabolism are also shown in Figure6A. For example, c-Jun is an example of a transcription factor subunit, marked by the AP-1 transcription factor, which is known to be involved in proliferation, differentiation, apoptosis and survival. HDAC4 and HDAC5 are histone deacetylases, which act as transcriptional repressors. Finally, c-SRC is a tyrosine protein kinase involved in cell-growth. Int. J. Mol. Sci. 2020, 21, 7814 12 of 23 Int. J. Mol. Sci. 2020, 21, x FOR PEER REVIEW 12 of 23

(A)

(B)

Figure 5. Cont. Int. J.Int. Mol. J. Mol. Sci. Sci.2020 2020, 21,, 21 7814, x FOR PEER REVIEW 13 of 2313 of 23

(C)

Figure 5. Gene regulatory networks stimulated by IL-2 (A), IL-7 (B), and IL-15 (C) in activated CD4+ T cells.Figure The 5. meaningGene regulatory of the icons networks/modules stimulated shown by in IL-2 this (A figure), IL-7 can(B), beand found IL-15 ( inC)the in activated software CD4 MetaCore.+ The IL-2T cells. network The meaning was clustered of the icons/ throughmodules the shown oncogene in this Src, figure general can be GPCR found signaling, in the software and the MetaCore transcription. factorThe AP-1. IL-2 Thenetwork IL-7 networkwas clustered shared through many similar the onco factorsgene asSrc, the general IL2 one, GPCR but it signaling, included andand expandedthe repertoiretranscription of nodes, factor specifically AP-1. The IL-7 around network the transcriptionalshared many similar regulators factors as (EGR1, the IL2 P73, one, and but it E2A). included The IL-15 networkand expanded appeared repertoire the most of dissimilar nodes, specifically to IL-2/7 around and radiated the transcripti throughonal a hubregulators of the (EGR1, transcription P73, and factors E2A). The IL-15 network appeared the most dissimilar to IL-2/7 and radiated through a hub of the FOXM1 and EPAS1 and cellular morphogenesis. transcription factors FOXM1 and EPAS1 and cellular morphogenesis. As compared to IL-2, the IL-7 network (Figure5B) shared many similar factors, but included and 2.6. Identify Genes Whose Expression Profiles Were Closely Correlated with the Growth Profiles of Activated expandedCD4+ Trepertoire Cells of nodes, specifically around the transcriptional regulators (EGR1, P73, and E2A). These regulators appeared to have obvious effects within the cell cycle (P21). In particular, EGR1 is The correlation between genes whose growth time profile matched the growth rate delivered for a zinc finger protein involved in transcriptional regulation and P73 is a tumor suppressor protein the respective experiment condition were conducted to identify the genes that may be of clinical value associated with P53. Conversely, the IL-15 network appeared the most dissimilar to IL-2/7 and radiated for genetic engineering of T cells. The T cell growth data and transcriptomic time profile data were throughtaken a for hub days of the 4, 5 transcription and 7 to determine factors whether FOXM1 there and EPAS1were changes and cellular to the morphogenesistranscriptome over (Figure the 5C). Specifically,course ofFOXM1 the experiment is a transcription and to compare factor to associatedthe growth profiles with cell of proliferation, activated CD4+ and T cells. EPAS1 The isgenes involved in theshowing induction the most of genes similar regulated time dynamics by oxygen. as the growth In addition rate are to found FOXM1 for andeach EPAS1, of the three other cytokines. transcription factorsUp-regulating are also found these importantgenes may inbe Figurehelpful5,to such promote as FOXO3A the growth (a transcriptionof activated CD4 factor+ T cells. associated Genes with apoptosis)with the and most PKC similar (a transcription dynamics as compared factor involved to the growth in cell adhesion).rate of the cell during IL-2 stimulation were analyzed. The same approach was taken for genes upregulated and matching the profile of Int. J. Mol. Sci. 2020, 21, x FOR PEER REVIEW 19 of 23

to determine the outliner genes shown in Figures 4A,C,E. In particular, the PCA approach published in our previous work [40–42] was implemented in this work.

Figure 6. An illustrative diagram of the fuzzy model to link the cytokine concentrations to the T cell growthFigure 6. rate. An Membershipillustrative diagram functions of werethe fuzzy used model to evaluate to link the the possibility cytokine concentr of each cytokineations to tothe be T lowcell orgrowth high (i.e.,rate.fuzzification), Membership functions which is followedwere used by to the evaluate linguistic the rulespossibility (i.e., inference) of each cytokine to determine to be low the growthor high rate(i.e., (i.e., fuzzification), defuzzification). which is followed by the linguistic rules (i.e., inference) to determine the growth rate (i.e., defuzzification). 2.6. Identify Genes Whose Expression Profiles Were Closely Correlated with the Growth Profiles of Activated CD4+ T Cells The Fuzzy modeling approach was selected over other models like simpler linear regression modelsThe due correlation to the following between reasons. genes whose Firstly, growth the fuzzy time model profile can matched return thelinguistic growth rules rate that delivered can be forunderstood the respective easily. experiment The inference condition system were in the conducted fuzzy approach to identify is the the significant genes that similarity may be ofto clinicalhuman valuelogic forand genetic decision-making, engineering as of the T cells. fuzzy The approach T cell growth can take data these and inconsiste transcriptomicncies into time account profile data and werehelp takencreate forinferences days 4, 5where and 7 a to human determine mind whether may fall there short. were In addition, changes tomachine the transcriptome learning algorithms over the coursefor Fuzzy of the models experiment have been and well to compare studied so to thethat growth the developed profiles model of activated has good CD4 prediction+ T cells. Thecapability. genes showingAs compared the most to a similarMamdani time Fuzzy dynamics Inference as the System growth th rateat is areadaptive, found linear for each regression of the three does cytokines. not take Up-regulatinginto account the these non-linearities, genes may be and helpful is not to easily promote adjustable the growth with of new activated information. CD4+ T cells. Genes with the mostFrom similar the growth dynamics data, as a compared Mamdani to fuzzy the growth model ratewas ofcreated the cell using during the IL-2 MATLAB stimulation command were analyzed.tunefis, whereby The same the approachFuzzy Interference was taken System for genes was upregulated developed using and matching the bounds the of profile the experiment of growth withfor each IL-7 of stimulation, the cytokines’ and concentrations. subsequently the In samethis instance, completed thewith bounds genes for upregulated the cytokine by concentrations IL-15. wereGenes 40 ng/mL, whose 100 growth ng/mL profiles and 100 most ng/mL similarly for IL-2, match IL-7 the and T cell IL-15, growth respectively. when stimulated After the by Fuzzy IL-2 includeInterference genes System such as: (FIS) TMEM74B was andcreated TTC36, and a transmembranethe membership protein functions and a tumorwere suppressor.established, IL-2the isparameters known to in be variables’ involved membership in many of thefunctions signaling were pathways tuned to minimizes involved in the growth root-mean and squared proliferation. error Paralleland get toa global growth minimum pathways, solution apoptosis using is closely the particle tied. Itswarm is not shocking,algorithm therefore[44–46]. that there would be genes involved in tumor suppression. Moving on to IL-7, genes whose growth profiles most similarly match the T cell growth when stimulated by IL-7 included ZNF665 and CA9, which are zinc finger proteins and zinc metalloenzymes, respectively. Zinc finger proteins have numerous functions in the human body, including participating in signal transduction, transcriptional regulation and other key metabolic processes. Additionally, another gene upregulated by IL-7 stimulation include ICAM4, which in involve intercellular adhesion. T cells normally form clusters with cells sticking together upon proliferation. Genes whose growth profiles most nearly matched the T cell growth when stimulated by IL-15 included genes such as PFKFB1, which is involved in glycolysis activation, and TPTE2, which is involved in inositol phospholipid metabolism, which would be used to form cell walls. The aforementioned genes which are similar to the growth rate of T cell stimulated solely with IL-2, IL-7 or IL-15, represent the polarizing conditions of the T cell, which would be the highest concentrations of IL-2, IL-7 and IL-15, respectively, which were compared to the control condition. This was done to see the numerical fold-increase from baseline for the time profile. Results from this study were also able to help delineate whether the genes were persistently upregulated compared to control, or changed the magnitude of the concentration as compared to control. Int. J. Mol. Sci. 2020, 21, 7814 15 of 23

3. Discussion

3.1. The Impact of Each Cytokine on T Cell Growths IL-2 and IL-7 had the largest impact on growth rates, however high levels of both did not result in the highest growth rates for this study. 15 ng/mL IL-2 with 6 ng/mL IL-7 (condition 200), 1 ng/mL IL-2 with 80 ng/mL IL-7 (condition 020), and 10 ng/mL IL-2 with 36 ng/mL IL-7 (condition 110) provided the best growth rates, and were statistically indistinguishable from each other. IL-15 was constant at 6 ng/mL for each of these top performers. Condition 011, corresponding to 1 ng/mL IL-2 and 36 ng/mL IL-7 and IL-15, was also among this statistical group of highly proliferative conditions. Much of the literature stipulated that there was a positive correlation between growth rate of CD4+ T cells and the presence of IL-2 and IL-7. The condition that would correlate with higher concentrations of these two cytokines is 110, which produces the third highest growth rate given the experimental data in Figure2A, and the highest growth rate in Figure1 after normalization. The conditions that produced the highest growth rate were 200 (highest IL-2) and 020 (high IL-7 and low IL-2 and -15) which was surprising. All conditions with IL-2, IL-7, and IL-15 present out-performed the control with none of these present; however moderate levels of each cytokine performed better than the extremes. The 000 condition (lowest non-zero concentrations of each cytokine) did not statistically perform any better than the control, and the 222 condition was very close to the average growth rate. This suggests that an optimal range for each cytokine exists, and over-stimulation by any cytokines may reduce proliferation of in-vitro activated CD4+ cultures with respect to the initial (seven-day), post-thaw cultures important for the CAR-T process.

3.2. The Factors that May Contribute to the Uncertainties in the Growth Rate Data After a third dataset was collected, we pooled the data for each of the conditions to demonstrate average findings over time between the data sets, presented in Figure1. The pooled data set does illustrate some uncertainties between the first two data sets and the average of all four data sets. There are a few possible reasons that this could be so, including: differentiation, donor specificity, T cell quantitation method, and nutrient exhaustion. Additionally, the cells were not co-activated with IL-2 until day 3 of the experiment, after a period of acclimation resulting from cryopreservation. The cells, although all from the same donor, may not have responded the same to cryopreservation, lending some uncertainty between data sets. Normalizing the data to the average growth rate of that set improved reliability of condition-to-condition comparisons. The first potential reason for the growth rate uncertainties is the possibility of cell differentiation. Activated T cells are undifferentiated to a certain extent, not yet at their final functional stage. It is known that cellular signaling pathways which are involved in proliferation are also intertwined with those involved in differentiation [2,8,24]. Another factor for the uncertainties shown in growth data is donor specificity. Donor specificity is always of interest in this type of experiment, because the consistency of T cells from an individual is a controlling factor in the consistency of the results. Patient to patient, there can be unexplained differences in how T cells react depending on gene expression in those specific patients. The data presented here represents two distinct donors, and initial viability from thaw was variable among experimental runs. It is a possibility that the cells could have been less responsive to the stimuli than in the previous run since the data points fall within a narrow standard deviation of each other. Another reason for the possible discrepancy between the data sets is the possibility of limitations of the method of quantitation, being that it was generated using hand-count techniques in hemocytometer. Using a hand-counting method would introduce potential variance and inconsistency in the measurement system via reproducibility, but also in repeatability when up against an analytical instrument. We did not conduct a gage R&R on the measurement system to confirm the assumption, but moving forward, verifying the accuracy and precision using a gage R&R would likely help remove human error as a factor in considering data uncertainty. Though, the TC20 Automated Cell Counter Int. J. Mol. Sci. 2020, 21, 7814 16 of 23 from Bio-Rad was unable to distinguish CD3/CD28 stimulation beads from activated CD4 T-cells, and gave unreliable results in the absence of stimulation beads due to the small diameter of recovering activated cells. This is why a hand-counting method was implemented in this work.

3.3. Discuss Major Findings in Transcriptomic Analysis (What Genes May Be of Clinical Value) There are some major findings in the transcriptomic data analysis. The first major finding was that many of the genes with persistent genes expressing changes in magnitude above control were associated with major pathways involved in cell growth and proliferation. This finding alone is helpful because the genes are markers that indicate how the cytokines impact the signaling within the cells. Metabolic pathways including: glycolysis and carbohydrate metabolism, lipid metabolism, mTOR, TCA cycle and PI3K-AKT are all pathways with which the up-and down-regulated genes were found to be associated, those of which are important for cell growth [22]. Genes specifically having to do with glycolysis would be a target of clinicians because glycolysis is necessary to fuel the cell during proliferation. Anabolism and cell division are costly processes to the cell, requiring a great deal of energy. With fuel for growth from glycolysis, PFKFB1 is an enzyme regulating the synthesis and degradation of fructose-2,6-bisphosphate. This gene was persistently upregulated in response to IL-15. Another gene of interest with clinical merit would be CXCL8, or interleukin-8 (IL-8). IL-8 is an inflammatory chemokine with neutrophil specificity, which can be seen as a chemo-attractant molecule [25]. From the outlier genes found from IL-2, IL-7 and IL-15, this outlier gene is common to both IL-2 and IL-7 stimulation in the transcriptomic data, and has association with many genes involved in TCR signaling. Another closely-related chemokine, CXCL9 was found to be upregulated by IL-2, IL-7 and IL-15. This molecule is of particular interest due to its involvement in immune cell migration, differentiation and activation [25]. Literature suggests that CXCL9 may be a target for cancer therapy as it is a tumor suppressor, and, coupled with other chemokines, can act as a potent differentiator to T helper cells in naïve T cells [26]. VEGFC, or vascular endothelial growth factor-C is another target gene of importance. It was upregulated upon IL-2 stimulation. The literature suggests that this compound has the ability to upregulate MAPK and AKT signaling pathways, both of which are direct links to cellular proliferation. Additionally, the VEGFC pathway is a pathway in itself, which is involved in directing growth of the lymphatic endothelial cells for transport throughout the body [27]. CCL2 was an outlier gene of the stimulation by IL-2. This gene is involved in T cell differentiation and is a monocyte chemo-attracting protein [28]. It could be a target of clinical value to preclude in the future for preventing differentiation.

3.4. The Complexity and Uncertainty of Naïve T Cell Growth, Activation and Pre-Differentiation Phenotyping This study focuses exclusively on the naïve CD4+ T cell phenotype. The cells were thawed and permitted to acclimate to the new post-thaw environment prior to activation and cytokine supplementation. Activation, however, does not consign differentiation, so although the naïve T cells were activated, they do not have to be referred to a specific T-cell subset. In a review published in Frontiers, the authors clearly state that, “after development in the thymus, conventional T cells exist as naïve, or undifferentiated cells, and upon activation can undergo clonal expansion and wield different effector functions,” [29]. Effector functions can serve to be a multitude of directions from a cellular cascade perspective. There are T-reg cells, Helper T Cells, Follicular Helper cells and an effector function can be a capability of cytotoxicity. Further, the authors state, “a trivial quantity of activated cells become memory cells,” which again implies that differentiation is downstream of activation [29]. Additionally, activation by cytokines and other chemokines differs drastically as compared to a specific antigen or pathogenic material. The most drastic difference noted by the authors in Frontiers is in the modulation of the survival programming, surrounding pathways such as apoptosis and necrosis [29]. Naïve T cells and their activated counterparts differ in this arena but it does not consign a different phenotype than the naïve phenotype initially. This is the main justification for using “activated” T cells, instead of “naïve” T cells throughout the paper. Int. J. Mol. Sci. 2020, 21, 7814 17 of 23

3.5. Design of Experiments The design of the experiments included, with the exception of the negative control, conditions including at least a low level of each cytokine in the mix, without single cytokine conditions. The four key reasons why single-cytokine groups were not implemented in this experiment include: first, IL-2, IL-7 and IL-15 all share the γ-chain in CD4+ lymphocytes, but the threshold for stimulation clearly is above the 000 level for all cytokines presented by the data and supported by literature. Additionally, it is also clear that a tertiary cocktail of all three γ-chain cytokines does not work in a synergistic way to raise the growth rate, as is evidenced by the “222” condition. Second, singular cytokines tend to have significant impacts by themselves with particular subsets of CD4+ T cells, i.e., IL-7 is pertinent for survival of memory T cell subsets, but not exclusively key for CD4+ naïve T cell phenotype expansion. The saturation of IL-7 in memory T cell survival is evidenced in the literature. Third, certain cytokines, such as IL-2 are known at higher concentrations to induce apoptosis, yet when coupled with other cytokines, there tends to be a synergistic effect [4]. Fourth and finally, the last reason we use combinations of cytokines is to reduce the concentrations of certain cytokines relative to one another. The impact of single cytokines can place cells down specific signaling pathways, which in reality is both unlikely to take place in vivo during lymphocyte expansion, and also is not likely to happen under normal cell conditions in humans. Many cytokines are present at once, all with their own individual concentrations; some are naturally occurring below 1ng/mL and others are present above 1ng/mL depending on the impact the cytokine has on a particular cell type. The cytokines we discuss here are pleiotropic in nature, so their impact specifically on CD4+ T cells is different than on other cell types.

3.6. Cytokine Storm, Cytokine Release Syndrome (CRS), and Relation to Immunotherapies, SARS-CoV-2 During the discovery phase of the testing of chimeric antigen receptor (CAR) T cells, one of the major findings was the consistent violent reaction of a patient’s own immune system to the engineered T cells [30]. History has shown that a patient’s own immune system can turn against the host when activated incorrectly by a pathogen or antigen-presenting cell (APC), such as during the 1918 Spanish Pandemic Flu. The severity of this pandemic was a result of the immune system’s unregulated cytokine release, which is termed a “cytokine storm.” Scientists later observed this phenomenon again during a disastrous failed Phase I clinical trial of CD-28 monocloncal antibody (mAb) superagonist TGN1412, where all six volunteers were plagued with cytokine storm shortly (minutes) after administration of the novel mAb [31]. Scientists have a better understanding of CRS and cytokine storm associated with especially the T cell immunotherapies now that there is a diverse set of data relating to different types of cancers as well as different demographics [32]. CD-19 blood-borne cancers like leukemia variants seem to be the best responders to CAR-T immunotherapies, especially in pediatric patients, who’s adaptive immune systems are underdeveloped and less coordinated [30]. Also, cancer load seems to play a large part in the severity of CRS for T cell immunotherapy patients’ immune systems to react proportionately [30]. Those with lower disease load or “burden” are less susceptible to CRS [31]. Although therapeutic drugs like Tocilizumab (IL-6 inhibitor) and Daclizumab (IL-2 inhibitor) have been shown to reduce CRS grade [31,32], new studies have shown, specifically that anti-inflammatory cytokines like IL-37 and IL-38 have inhibitive effects on IL-1, which in the recent SARS CoV-2 (COVID-19) pandemic outbreak has played a part in our understanding of cytokine release and cytokine-related inflammation [33]. CAR-T therapies also have an inductive effect on IL-1, the pro-inflammatory cytokine, which as previously mentioned, can potentially be neutralized using the anti-inflammatory effect of IL-37 [34]. Finally, one of the root sources of the IL-1 induction by SARS-CoV-2 seems to be mast cell activation by the virus itself [35]. The mast cell is a histamine-producing part of the innate immune system, rather than the adaptive immune system with T cells, however, recognition of pathogens by these mast cells ultimately impacts the adaptive immune system through cytokine signaling. Using anti-inflammatory cytokine signaling, there is a possibility to reduce the high levels of inflammation seen in CAR-T immunotherapies. Int. J. Mol. Sci. 2020, 21, 7814 18 of 23

3.7. Limitation of This Work There are a few limiting factors of this work. Given there was a shortage of cells for the experiments, triplicates of each of the 16 experimental conditions were not available. The shortage of cells created uncertainty in the data sets both from a replicability, and also reproducibility perspective. Additionally, and perhaps of equal relevance, the RNA sequencing data that was able to be gathered was not complete due to the limited time and available resources, in the sense that only the effects of individual cytokines were able to be analyzed. There were no coupling or interactive effects that could be studied. This has particular relevance in that the conditions identified as most proliferative in the study were not able to be evaluated because cytokine coupling was not part of the testing completed. Suggestions for future work would be several recommendations. The first recommendation would be to take measures to minimize the uncertainty in the data sets. The uncertainty was explained, however maintaining the same donor throughout the experiment sets would be one suggestion to reduce variance. In addition, verifying the measurement system from the perspective of replicability versus reproducibility is especially important in developing consistent results. Other T-cell phenotypes will also be investigated, and preliminary data has been collected for CD8 subsets, as they tend to be favored for CAR-T processes currently. Finally, RNA Sequencing should be conducted for more experimental conditions, especially those conditions with various cytokine combinations.

4. Materials and Methods

4.1. Cell Culture and Cytokine Stimulation 1-mL vials of naïve CD4+ T-cells from a human donor were purchased (Astarte Biologics, Bothell, WA, USA) and stored in the vapor phase of liquid nitrogen until ready for culture. The cells were thawed in a 37 ◦C water bath for 1–2 min or until pipettable, transferred to a 15-mL conical tube, and resuspended in 9 mL of 37 ◦C X-VIVO 15 serum-free cell culture medium (Lonza, Basel, Switzerland). The cells were pelleted via centrifugation at 200 g for 10 min, the supernatant was × removed, and the pellet was resuspended in 10 mL of 37 ◦C culture medium before determining live cell concentration with Trypan Blue (Gibco Laboratories, Gaithersburg, MD, USA) and manual hemocytometer counts. The cell concentration was adjusted to 166,667 cells/mL with additional 37 ◦C medium to achieve desired seeding number (50,000 cells/well) before addition of Dynabeads Human T-Activator CD3/CD28 activation beads (Gibco Laboratories) in a 1:1 (bead:cell) ratio. The solution was then plated in 300 µL volumes on two round-bottom polypropylene 96-well plates from Corning Life Sciences (Corning, NY, USA) and incubated at 37 ◦C for three days.

4.2. Growth Plate After the three day post-thaw acclimation, recombinant human IL-2, IL-7, and IL-15, sourced from Miltenyi Biotec (Bergisch Gladbach, Germany) and diluted in sterile-filtered water, were pipetted into individual wells at the 16 experimental concentrations below to give quadruplicate wells per cytokine condition. The key index, along with the actual concentration of each cytokine is listed in Table2. One well of each condition was sampled on days 3, 5, and 6 via Trypan Blue exclusion and hemocytometer count to determine cell concentration, and every well was sampled on days 4 and 7. Growth rates were calculated between days three and seven except for the third data set, where day six was chosen as the endpoint. In that experiment, many of the conditions reached prohibitively dense populations (approaching or exceeding 106 cells/mL) by day six, and consistently declined afterward to day seven, suggesting depletion of medium nutrients or other limitations of overly-confluent cultures. The first three days of the culture were not included in calculations as growth is stagnated during this period due to media acclimation and post-activation anergy. Simple exponential growth was assumed and growth rates were calculated as: c2 ln c G = 1 (1) (t t ) 2 − 1 Int. J. Mol. Sci. 2020, 21, 7814 19 of 23

where c1 and c2 are live cell concentrations at times t1 and t2 respectively. The growth rates of each condition were then normalized to the average growth rate of that respective data set, as there were inter-experiment diﬀerences in maximum cell concentrations (please refer to Section 3.2).

4.3. Transcriptomics Analysis Eighteen wells each of conditions 000, 200, 020, and 002 (as shown in Table2) were replicated on the second 96-well plate for RNA extraction at three time points (days 4, 5 and 7). At each time point 6 wells for each of these conditions were pooled and the cells lysed with a RNeasy Plus Mini kit (Qiagen, Hilden, Germany). The lysates were frozen at 70 C until RNA isolation could be performed. − ◦ Sequencing of the isolated total RNA was performed by the Center for Medical Genomics (Indianapolis, IN, USA). The following protocol was used for RNA seq analysis. After obtaining the FASTQ data, it had to be processed and examined. In order to process the FASTQ data, alignment software was used. STAR software was the chosen software. STAR stands for Spliced Transcripts Alignment to a Reference, which is a novel software RNA-seq alignment algorithm that uses “sequential maximum mappable seed search in uncompressed suffix arrays followed by seed clustering and stitching procedure” [36]. Subsequently, NGSUTils software suite was used. NGSUTils is a software tool suite for examining data from RNA Sequencing experiments, such as FASTQ files, which is what was used as the initial reads from the RNA Sequencing data from this experiment [37]. featureCounts is a software first introduced in Bioinformatics. The software is a read summarization program used for counting reads produced from RNA Sequencing tests. The highlight from featureCounts is that it utilizes chromosome hashing and feature blocking techniques. The software works with single and paired-end reads [38]. Finally, edgeR was used to analyze the gene expression data and perform several analytical functions to examine the count data [39]. On the basis of the expression profiles returned by edgeR, principal component analysis (PCA), one of the most popular approaches to project high-dimensional data into a two-dimensional space, was used to determine the outliner genes shown in Figure4A,C,E. In particular, the PCA approach published in our previous work [40–42] was implemented in this work.

4.4. Fuzzy Modeling On the basis of the growth data, a Fuzzy model was developed to predict the growth rate of activated CD4+ T cells stimulated by IL-2, IL-7 and IL-15. In the fuzzy model, the value of each of the inputs (i.e., concentrations of IL-2, IL-7, and IL-15) and the output (i.e., the growth rate of activated CD4+ T cells) was categorized into Low or High label (to reduce the number of model parameters). The growth data was then fed into the Fuzzy model so that an Adaptive Neuro Fuzzy Inference System [32–34,43] was to build the linguistic rules to link the inputs and output. One example linguistic rule is “If IL-2 concentration is high, IL-7 concentration is low, and IL-15 concentration is low, the growth rate is high”. A membership function (e.g., Gaussian function) was assigned for each linguistic label (e.g., Low) of each variable (i.e., three inputs and output) to quantify the probability of that variable falling into each category. Then Fuzzy model then infers the value of the output on the basis of the variables’ membership functions and linguistic rules. Figure6 shows an illustrative diagram of the Fuzzy model. The Fuzzy modeling approach was selected over other models like simpler linear regression models due to the following reasons. Firstly, the fuzzy model can return linguistic rules that can be understood easily. The inference system in the fuzzy approach is the signiﬁcant similarity to human logic and decision-making, as the fuzzy approach can take these inconsistencies into account and help create inferences where a human mind may fall short. In addition, machine learning algorithms for Fuzzy models have been well studied so that the developed model has good prediction capability. As compared to a Mamdani Fuzzy Inference System that is adaptive, linear regression does not take into account the non-linearities, and is not easily adjustable with new information. Int. J. Mol. Sci. 2020, 21, 7814 20 of 23

From the growth data, a Mamdani fuzzy model was created using the MATLAB command tuneﬁs, whereby the Fuzzy Interference System was developed using the bounds of the experiment for each of the cytokines’ concentrations. In this instance, the bounds for the cytokine concentrations were 40 ng/mL, 100 ng/mL and 100 ng/mL for IL-2, IL-7 and IL-15, respectively. After the Fuzzy Interference System (FIS) was created and the membership functions were established, the parameters in variables’ membership functions were tuned to minimizes the root-mean squared error and get a global minimum solution using the particle swarm algorithm [44–46]. After the rules were established and membership functions were estimated, the Fuzzy model was evaluated by testing the growth data against the model for eight new conditions (inputs). After the Fuzzy model was validated, in order to test all possible combinations of cytokines between the aforementioned bounds set forth, a 3-D surface scatter heat map was developed to predict the conditions that would produce the highest growth rate of the T cells from the Fuzzy model. In Figure3, the bright yellow color represents the high growth conditions while the blue represents the low growth conditions.

5. Conclusions To conclude, we have investigated the growth of activated CD4+ T cells via the influence of three key cytokines: IL-2, IL-7 and IL-15. Through a design of experiments with different cytokine concentrations, we identified several key conditions to optimize the growth of the activated CD4+ T cells. The most favorable conditions for T cell growth study, and from the predictive model align to include 10 ng/mL for IL-2 and 36 ng/mL for IL-15, followed by 36 ng/mL for IL-7 and 36 ng/mL for IL-15, followed by 10 ng/mL of IL-2 and 36 ng/mL for IL-15. These conditions led to the highest growth rate of activated CD4+ T cells in our study with statistical significance. The Fuzzy Inference System successfully analyzed the growth of the cells under the influence of the cytokines, and to predicts moderate concentrations of synergetic cytokine cocktails to optimize the growth of the cells. To explore and support the result, we also utilized RNA Sequencing analysis to take several gene expression snapshots to understand the impact of each cytokine on the genes up- and down-regulated from the cells during the experiment. We found genes relating to different signaling pathways associated with T cell growth, as well as specific signaling pathways that each cytokine influenced in the T cells using principal component analysis, GeneMANIA and MetaCore. These signaling pathway include TCR, JAK/STAT, MAPK, AKT and PI3K-AKT signaling. In particular, IL-2 regulated genes involved in cell and leukocyte chemotaxis, chemokine receptor activity, cellular defense response and cell motility. IL-7 regulated the genes associated with chloride transport, cell chemotaxis, chloride transmembrane activity, and cytokine receptor activity, and IL-15 regulated genes for cell growth, zinc ion responses, cellular response to metal ions, and regulation of calcium ion transport into the cell. Lastly, growth data analysis indicated a statistically significant interaction between IL-2 and IL-7, reproduced in all data sets and lays the groundwork for expositing the combination of these two cytokines and how they act synergistically together for the CD4+ T cell phenotype. We recommend a system for clinicians to utilize to accelerate the growth of initially naïve CD4+ T cells and lay the groundwork for an approach to identifying optimized growth potential in other types of human leukocytes.

Supplementary Materials: Supplementary Materials can be found at http://www.mdpi.com/1422-0067/21/21/ 7814/s1. Author Contributions: Conceptualization of experimental design and analysis of results, W.J.K. and C.C.; Analysis of Transcriptomic data sets, B.H., S.H. and Z.H.; Modeling and Simulation for ANFIS, B.H. and Z.H.; Writing—original draft preparation, B.H. and C.C.; funding acquisition, W.J.K. and Z.H. Writing—review & editing, Z.D., Z.H. and W.J.K. All authors have read and agreed to the published version of the manuscript. Funding: This material is based in part upon work supported by the National Science Foundation under grant no. NSF:EAGER 1645225. This work was also performed under a Project Award Agreement from the National Institute for Innovation in Manufacturing Biopharmaceuticals (NIIMBL) and financial assistance award 70NANB17H002 from the U.S. Department of Commerce, National Institute of Standards and Technology. Conflicts of Interest: The authors declare no conflict of interest. Int. J. Mol. Sci. 2020, 21, 7814 21 of 23

References

1. Das, K.R.; Vernau, L.; Grupp, S.A.; Barrett, D.M. Naive T-cell Deficits at Diagnosis and after Chemotherapy Impair Cell Therapy Potential in Pediatric Cancers. Cancer Discov. 2019, 9, 492–499. [CrossRef][PubMed] 2. Read, K.A.; Powell, M.D.; McDonald, P.W.; Oestreich, K.J. IL-2, IL-7 and IL-15: Multistage regulators of CD4+ T helper cell differentiation. Exp. Hematol. 2016, 44, 799–808. [CrossRef][PubMed] 3. Geginat, J.; Sallusto, F.; Lanzavecchia, A. Cytokine-driven Proliferation and Differentiation of Human Naive, Central Memory, and Effector Memory CD4+ T Cells. J. Exp. Med. 2001, 194, 1711–1720. [CrossRef][PubMed] 4. Jaleco, S.; Swainson, L.; Dardalhon, V.; Burjanadze, M.; Kinet, S.; Taylor, N. Homeostasis of Naive and Memory CD4+ T Cells: IL-2 and IL-7 Differentially Regulate the Balance Between Proliferation and Fas-Mediated Apoptosis. J. Immunol. 2003, 171, 61–68. [CrossRef][PubMed] 5. Alves, L.N.; Hooibrink, B.; Arosa, F.A.; van Lier, R.A.W. IL-15 induces antigen-independent expansion and differentiation of human naive CD8+ T cells in vitro. Blood 2003, 102, 2541–2546. [CrossRef] 6. Unutmaz, D.; Pileri, P.; Abrignani, S. Antigen-Independent Activation of Naive and Memory Resting T Cells by a Cytokine Combination. J. Exp. Med. 1994, 180, 1159–1164. [CrossRef] 7. Kanegane, H.; Tosato, G. Activation of Naive and Memory T Cells by Interleukin-15. Blood 1996, 88, 230–235. [CrossRef] 8. Golubovskaya, V.;Wu, L. Different Subsets of T Cells, Memory,Effector Functions, and CAR-T Immunotherapy. Cancers 2016, 8, 36. [CrossRef] 9. Sommermeyer, D.; Hudacek, M.; Kosasih, P.L.; Gogishvili, T.; Maloney, D.G.; Turtle, C.J.; Riddell, S.R. Chimeric antigen receptor-modified T cells derived from defined CD8+ and CD4+ subsets confer superior antitumor reactivity in vivo. Leukemia 2016, 30, 492–500. [CrossRef][PubMed] 10. Wang, D.; Aguilar, B.; Starr, R.; Alizadeh, D.; Brito, A.; Sarkissian, A.; Ostberg, J.R.; Forman, S.J.; Brown, C.E. Glioblastoma-targeted CD4+ CAR T cells mediate superior antitumor activity. JCI Insight 2018, 3.[CrossRef] 11. Litvinova, L.S.; Sokhonevich, N.A.; Gutsol, A.A.; Kofanova, K.A. The Influence of Immunoregulatory Cytokines IL-2, IL-7, and IL-15 upon Activation, Proliferation, and Apoptosis of Immune Memory T-cells in vitro. Cell Tissue Biol. 2013, 7, 539–544. [CrossRef] 12. Silva, S.L.; Albuquerque, A.S.; Matoso, P.; Charmeteau-de-Muylder, B.; Cheynier, R.; Ligeiro, D.; Abecasis, M.; Anjos, R.; Barata, J.T.; Victorino, R.M.; et al. IL-7-Induced Proliferation of Human Naive CD4 T-Cells Relies on Continued Thymic Activity. Front. Immunol. 2017, 8, 20. [CrossRef][PubMed] 13. Dooms, H.; Abbas, A.K. Control of CD4+ T-cell memory by cytokines and costimulators. Immunol. Rev. 2006, 211, 23–38. [CrossRef][PubMed] 14. Kinter, A.L.; Godbout, E.J.; McNally, J.P.; Sereti, I.; Roby, G.A.; O’Shea, M.A.; Fauci, A.S. The Common Gamma-Chain Cytokines IL-2, IL-7, IL-15, and IL-21 Induce the Expression of Programmed Death-1 and Its Ligands. J. Immunol. 2008, 181, 6738–6746. [CrossRef] 15. Fisher, R.A. 014: On the ‘Probable Error’ of a Coefficient of a Correlation Deduced from a Small Sample. Metron 1921, 1, 1–32. 16. Mamdani, E.H.; Assilian, S. An Experiment in Linguistic Synthesis with a Fuzzy Logic Controller. Int. J. Man-Mach. Stud. 1975, 7, 1–13. [CrossRef] 17. Wold, S.; Esbensen, K.; Geladi, P. Principal Component Analysis. Chemom. Intell. Lab. Syst. 1987, 2, 37–52. [CrossRef] 18. Warde-Farley, D.; Donaldson, S.L.; Comes, O.; Zuberi, K.; Badrawi, R.; Chao, P.; Franz, M.; Grouios, C.; Kazi, F.; Lopes, C.T.; et al. The GeneMANIA Prediction server: Biological network integration for gene prioritization and predicting gene function. Nucleic Acids Res. 2010, 38, W214–W220. [CrossRef] 19. Dubovenko, A.; Nikolsky, Y.; Rakhmatulin, E.; Nikolskaya, T. Functional Analysis of OMICs Data and Small Molecule Compounds in an Integrated Knowledge-Based Platform. In Biological Networks and Pathway Analysis; Humana Press: New York, NY, USA, 2017. 20. TCR Signaling in Naive CD4 T Cells. NDEx. Available online: http://www.ndexbio.org/#/network/0c2862fa- 6196-11e5-8ac5-06603eb7f303 (accessed on 24 August 2020). 21. Tokunaga, R.; Zhang, W.; Naseem, M.; Puccini, A.; Berger, M.D.; Soni, S.; McSkane, M.; Baba, H.; Lenz, H.J. CXCL9. CXCL10, CXCL11/CXCR3 axis for immune activation—A target for novel cancer therapy. Cancer Treat. Rev. 2018, 63, 40–47. [CrossRef] Int. J. Mol. Sci. 2020, 21, 7814 22 of 23

22. Chien, M.; Ku, C.; Johansson, G.; Chen, M.; Hsiao, M.; Su, J.; Inoue, H.; Hua, K.; Wei, L.; Kuo, M. Vascular Endothelial Growth Factor-C (VEGF-C) promotes angiogenesis by induction of COX-2 in leukemic cells via the VEGF-R3/JNK/AP-1 pathway. Carcinogenesis 2009, 30, 2005–2013. [CrossRef] 23. Tacconi, C.; Ungaro, F.; Correale, C.; Arena, V.; Massimino, L.; Detmar, M.; Spinelli, A.; Carvello, M.; Mazzone, M.; Oliveira, A.I.; et al. Activation of the VEGFC/VEGFR3 Pathway Induces Tumor Immune Escape in Colorectal Cancer. Cancer Res. 2019, 79, 4196–4210. [CrossRef][PubMed] 24. Rochman, Y.; Spolski, R.; Leonard, W.J. New Insights into the regulation of T cells by γc Family Cytokines. Nat. Rev. Immunol. 2009, 9, 480–490. [CrossRef][PubMed] 25. Krebs, R.; Tikkanen, J.M.; Ropponen, J.O.; Jeltschm, M.; Jokinen, J.J.; Yla-Herttuala, S.; Nykanen, A.I.; Lemstrom, K.B. Critical Role of VEGF-C/VEGFR-3 Signaling in Innate and Adaptive Immune Responses in Experimental Obliterative Bronchiolitis. Am. J. Pathol. 2012, 181, 1607–1620. [CrossRef][PubMed] 26. Mason, E.F.; Rathmell, J.C. Cell metabolism: An essential link between growth and apoptosis. Biochim. Biophys. Acta 2011, 1813, 645–654. [CrossRef][PubMed] 27. Rauniyar, K.; Jha, S.J.; Jeltsch, M. Biology of Vascular Endothelial Growth Factor C in the Morphogenesis of Lymphatic Vessels. Front. Bioeng. Biotechnol. 2018, 6, 7. [CrossRef][PubMed] 28. Luther, S.A.; Cyster, J.G. Chemokines as regulators of T cell differentiation. Nat. Immunol. 2001, 2, 102–107. [CrossRef][PubMed] 29. Zhan, Y.; Carrington, E.M.; Zhang, Y.; Heinzel, S.; Lew, A.M. Life and Death of Activated T Cells: How Are They Different From Naïve T Cells? Front. Immunol. 2017, 8, 1809. [CrossRef] 30. Lee, D.W.; Gardner, R.; Porter, D.L.; Louis, C.U.; Ahmed, N.; Jensen, M.; Grupp, S.A.; Mackall, C.L. Current concepts in the diagnosis and management of cytokine release syndrome. Blood 2014, 124, 188–195. [CrossRef] 31. Suntharalingam, G.; Perry, M.R.; Ward, S.; Brett, S.J.; Castello-Cortes, A.; Brunner, M.D.; Panoskaltsis, N. Cytokine storm in a phase 1 trial of the anti-CD28 monoclonal antibody TGN1412. N. Engl. J. Med. 2006, 355, 1018–1028. [CrossRef] 32. Hay, K.A.; Hanafi, L.A.; Li, D.; Gust, J.; Liles, W.C.; Wurfel, M.M.; López, J.A.; Chen, J.; Chung, D.; Harju-Baker, S.; et al. Kinetics and biomarkers of severe cytokine release syndrome after CD19 chimericantigen receptor-modified T-cell therapy. Blood 2017, 130, 2295–2306. [CrossRef] 33. Conti, P.; Ronconi, G.; Caraffa, A.; Gallenga, C.E.; Ross, R.; Frydas, I.; Kritas, S.K. Induction of Pro-inflammatory Cytokines (IL-1 and IL-6) and Lung Inflammation by Coronavirus-19 (COVI-19 or SARS-CoV-2): Anti-inflammation Strategies. J. Biol. Regul. Homeost. Agents 2020, 34, 1. 34. Caraffa, A.; Gallenga, C.E.; Kritas, S.K.; Ronconi, G.; Di Emidio, P.; Conti, P. CAR-T Cell Therapy Causes Inflammation by IL-1 which Activates Inflammatory Cytokine Mast Cells: Anti-inflammatory Role of IL-37. J. Biol. Regul. Homeost. Agents 2019, 33, 1671–1674. 35. Kritas, S.K.; Ronconi, G.; Caraffa, A.; Gallenga, C.E.; Ross, R.; Conti, P. Mast Cells Contribute to Coronavirus-induced Inflammation: New Anti-inflammatory Strategy. J. Biol. Regul. Homeost. Agents 2020, 34, 9–14. [PubMed] 36. Dobin, A.; Davis, C.A.; Schlesinger, F.; Drenkow, J.; Zaleski, C.; Jha, S.; Batut, P.; Chaisson, M.; Gingeras, T.R. STAR: Ultrafast Universal Rna-Seq Aligner. Bioinformatics 2012, 29, 15–21. [CrossRef] 37. Breese, M.R.; Liu, Y. NGSUtils: A Software Suite for Analyzing and Manipulating Next-Generation Sequencing Datasets. Bioinformatics 2013, 29, 494–496. [CrossRef][PubMed] 38. Liao, Y.; Smyth, G.K.; Shi, W. FeatureCounts: An Efficient General Purpose Program for Assigning Sequence Reads to Genomic Features. Bioinformatics 2013, 30, 923–930. [CrossRef] 39. Robinson, M.D.; McCarthy, D.J.; Smyth, G.K. edgeR: A Bioconductor package for differential expression analysis of digital gene expression data. Bioinformatics 2010, 26, 139–140. [CrossRef][PubMed] 40. Hua, M.; Huang, W.; Chen, A.; Rehmet, M.; Jin, C.; Huang, Z. Comparison of Antimicrobial Resistance Detected in Environmental and Clinical Isolates from Historical Data for the US. BioMed Res. Int. 2020, 2020, 4254530. [CrossRef][PubMed] 41. Yang, K.; Wang, A.; Fu, M.; Wang, A.; Chen, K.; Jia, Q.; Huang, Z. Investigation of Incidents and Trends of Antimicrobial Resistance in Foodborne Pathogens in Eight Countries from Historical Sample Data. Int. J. Environ. Res. Public Health 2020, 17, 472. [CrossRef] 42. Shing, J.; Jang, R. ANFIS: Adaptive-Network-Based Fuzzy Inference System. IEEE Trans. Syst. Man Cybern. 1993, 23, 665–685. Int. J. Mol. Sci. 2020, 21, 7814 23 of 23

43. Teachey, D.T.; Lacey, S.F.; Shaw, P.A.; Melenhorst, J.J.; Maude, S.L.; Frey, N.; Pequignot, E.; Gonzalez, V.E.; Chen, F.; Finklestein, J.; et al. Identiﬁcation of Predictive Biomarkers for Cytokine Release Syndrome after Chimeric Antigen Receptor T-cell Therapy for Acute Lymphoblastic Leukemia. Cancer Discov. 2016, 6, 664–679. [CrossRef][PubMed] 44. Ghanbari, A.; Ghaderi, S.F.; Azadeh, M.A. Adaptive Neuro-Fuzzy Inference System vs. Regression Based Approaches for Annual Electricity Load Forecasting. In Proceedings of the 2nd International Conference on Computer and Automation Engineering, Singapore, 26–28 February 2010. 45. Esmaeli, M.; Osanloo, M.; Rashidinejad, F.; Bazzazi, A.A.; Taji, F. Multiple regression, ANN and ANFIS models for prediction of backbreak in the open pit blasting. Eng. Comput. 2014, 30, 549–558. [CrossRef] 46. Mokarram, M.; Amin, H.; Khosravi, M.R. Using Adaptive Neuro-Fuzzy Inference System and Multiple Linear Regression to Estimate Orange Taste. Food Sci. Nutr. 2019, 7, 3176–3184. [CrossRef][PubMed]

Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional aﬃliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).